Cardiac MRI Through-Plane Super-Resolution Guided by Reference and Memory

2026-07-08

AI Summary

Clinical cardiac MRI suffers from low through-plane resolution due to constraints imposed by scan duration and physiological motion, which hinders accurate three-dimensional analysis and diagnosis. To address this limitation, this work proposes the STRMSR framework, which leverages a high-resolution reference view from the same subject as external memory and dynamically constructs an intermediate memory bank using successive super-resolution outputs. A coarse-to-fine context matching mechanism guides the reconstruction process, while a novel learnable block-wise dynamic feature fusion module ensures inter-slice consistency and enables high-quality through-plane super-resolution. Evaluated on the WHS dataset under both orthogonal-plane and long-axis chamber reference protocols, the method consistently outperforms existing baselines at both 4× and 8× super-resolution factors.

Abstract

Clinical cardiac MRI is commonly acquired with high in-plane resolution but coarse through-plane resolution to reduce scan time and accommodate breath-hold and cardiac-motion constraints, which limits 3D analysis and diagnostic accuracy. We propose STRMSR, a reference- and memory-guided through-plane super-resolution (SR) framework that reconstructs high-resolution (HR) cardiac volumes by leveraging HR reference views acquired from the same subject and intermediate SR results as the memory. Our method uses coarse-to-fine contextual matching to establish robust correspondence between low-resolution target and reference/memory images under spatial misalignment. A learnable patch-wise dynamic feature aggregation module predicts content-adaptive mixture weights for each local patch, effectively fusing dynamic information while suppressing unreliable feature transfers. The intermediate SR results stored in the memory bank ensure slice-to-slice consistency for the super-resolved 3D volume. Experiments on the WHS cardiac MRI dataset under two reference protocols, orthogonal-plane views and long-axis chamber views, demonstrate consistent improvements over baselines at 4x and 8x upsampling factors.

Problem

• cardiac MRI
• through-plane super-resolution
• low through-plane resolution
• 3D analysis
• diagnostic accuracy

Innovation

• through-plane super-resolution
• reference-guided reconstruction
• memory-augmented MRI
• dynamic feature aggregation
• cardiac MRI